Content Labelling of Hidden Services With Keyword Extraction Using the Graph Decomposition Method

Abstract

The data investigation of hidden services on the dark web is gaining attention from the research community and law enforcement agencies. However, the anonymity feature of hidden services makes it difficult to index the hidden services for investigation. Therefore, one of the primary focuses of dark web data investigation research is labelling the hidden services so that the labelled services can be classified or indexed further. The methodology deployed in the proposed work is based on keyword extraction using the graph degeneracy method. The proposed work analyzes the text data by extracting keywords from each hidden service document. The accuracy of the proposed method is validated by LDA-based topic modelling approach. The document labelling obtained by the keyword extraction method and LDA model matched with the accuracy of 78. The main intuition behind the keywords extraction method is that central nodes make good keywords. This is because central nodes with high centrality in the GoW of a document correspond to the document's keywords, which are well-understood by humans.